KM3BUU

The KM3BUU project is an integrated environment for the GiBUU studies within the KM3NeT experiment.

Installation

The main KM3BUU project is python based using singularity for running GiBUU. The python environment is required to have version 3.5 or higher and singularity is required to have version 3.3 or higher (e.g. v3.4). In the default workflow the singularity image is build remote from the KM3NeT docker server, but it can also be built locally (see :ref:`Local Machine`). KM3BUU is not provided via python package manager and can be installed as follows. First the repository needs to be cloned:

git clone https://git.km3net.de/simulation/km3buu
cd km3buu

After downloading the repository the package can be installed via:

pip install -e .

GiBUU Only Usage

The repository design also allows the usage without python environment. In this scenario the singularity container containing the GiBUU environment has to be built first. This can be done locally if root privileges are available:

make build

If root privileges are not available, e.g. running the KM3BUU on a compute cluster, it also can be done remote via the KM3NeT docker server:

make buildremote

If the python environment is used afterwards, the file path of the container can be written to the configuration file and is not required to be built again.

For running GiBUU the used jobcards have to be moved to a sub-folder within the jobcards folder of the project. Each sub-folder represents a set of jobcards, which can be passed to GiBUU by:

make run CARDSET=examples

This specific command runs all jobcards within the jobcards/examples folder and stores the output inside the folder output. The folder structure is applied from the jobcardsfolder.

Theory

In order to retrieve correct results and provide correct KM3NeT weights (w2) the treatment of the GiBUU weights is an important step. A brief description of the GiBUU weights and how to calculate actual cross sections is given on the GiBUU Homepage and a more detailed description of the calculation can be found in the PhD Thesis of Tina Leitner in Chapter 8.3. As it is mentioned in the description of the output flux file in the documentation this is not taken somehow into account inside the weights. Following the description the GiBUU event weight can be converted to a binned cross section via

\frac{d\sigma}{E} = \frac{\sum_{i\in I_\text{bin}} w_i}{\Delta E}\cdot\frac{1}{E\Phi},

where :math:`\Phi`__ is the simulated flux As the weights are given for each run individually the weight also has to be divided by the number of runs.

Tutorial

The python framework is build around the GiBUU workflow, i.e. a jobcard is processed and the output files are written out. The jobcards are technically FORTRAN namelists and can be created using a Jobcard object. In the example this is done via loading an existing jobcard:

>>> from km3buu.jobcard import Jobcard, read_jobcard

>>> jc = read_jobcard("jobcards/examples/example.job")

In the next step the jobcard is processed:

>>> from km3buu.ctrl import run_jobcard

>>> run_jobcard(jc, "./output")
0

Finally, the output can be parsed using a GiBUUOutput object:

>>> from km3buu.output import GiBUUOutput

>>> data = GiBUUOutput("./output")